我使用for循环来计算y_pred和y_true之间的精度，这使我的代码非常慢，因此我想避免这种for循环。如何更改此代码？

sumacc =  tf.Variable(0,dtype=tf.float32)
for i in range(0,y_pred_batch_size):
    ele_y_pred = y_pred[i,0:label_length[i,0]]
    ele_y_true = y_true[i,0]]
    sumacc =  sumacc + tf.reduce_mean(tf.cast(tf.equal( ele_y_pred,ele_y_true),tf.float32)) 
tf.print("\n","accuracy: ",sumacc / tf.constant(y_pred_batch_size,dtype=tf.float32) * 100,"%" )

y_pred和y_true的值看起来像.. [batch_index,[[45],[56],[12],...]